Coffee production process parameter optimization method, system, equipment and medium

Abstract

The invention discloses a method, a system, equipment and a medium for optimizing coffee production process parameters, and relates to the technical field of food industry data analysis. Receiving a flavor target and a brewing mode of the coffee beans to be roasted by a user; processing the flavor target based on a preset coffee flavor dimension table to obtain a target flavor portrait; correcting the target flavor portrait based on a flavor adjustment mode of a brewing mode to obtain a final flavor portrait; inputting the physical parameters and the final flavor portraits into a model to obtain a plurality of predicted flavor portraits; and calculating each predicted flavor portrait and the final flavor portrait, selecting the candidate process parameter corresponding to the maximum value from the plurality of quality similarity scores, determining the candidate process parameter as a baseline optimal process parameter, and sending the baseline optimal process parameter to the first user so that the first user can check the optimal process parameter. By implementing the provided technical scheme, the limitation of a traditional static preset baking curve is broken through, and the optimal process parameters can be dynamically generated according to specific raw material characteristics and user requirements.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of food industry data analysis, in particular to a coffee production process parameter optimization method, system, device and medium. BACKGROUND

[0002] With the continuous maturation and upgrading of the global coffee consumption market, consumers' pursuit of coffee flavor and quality has become increasingly sophisticated and personalized. Coffee roasting, as a key transformation link from green beans to finished coffee, the precise control and optimization of its production process directly determines the final flavor characteristics, aroma complexity and taste balance of coffee, and has become the core driving force for the development of coffee industry technology.

[0003] Existing coffee roasting technology has achieved a certain degree of automation control. This type of technology is usually applied in roasting equipment integrated with programmable controllers. The equipment can automatically control the fire, damper and other actuators during the roasting process according to the pre-stored roasting curve program. By collecting real-time bean temperature with temperature sensors, the equipment system strives to make the actual temperature curve as close as possible to the preset target curve to complete standardized roasting operations.

[0004] However, the preset roasting curve relied on by the above-mentioned coffee roasting method is static and fixed, and does not take into account the physical property differences of coffee beans and the brewing method, making it difficult to dynamically generate an optimal process parameter scheme for specific needs and raw materials, resulting in difficulty in accurately matching product flavor to user expectations and affecting user experience. SUMMARY

[0005] The present application provides a coffee production process parameter optimization method, system, device and medium, which breaks through the limitations of traditional static preset roasting curve, dynamically generates optimal process parameters according to specific raw material characteristics and user needs, more accurately matches user expected flavor characteristics, and improves user experience.

[0006] In a first aspect, the application provides a method for optimizing coffee production process parameters, the method comprising: receiving a roasting requirement of a first user for coffee beans to be roasted, analyzing the roasting requirement to obtain a flavor target and a brewing method; quantifying the flavor target into a set of flavor indicators and corresponding flavor scores based on a preset coffee flavor dimension table to form a target flavor profile; finding a flavor adjustment method corresponding to the brewing method from a preset brewing flavor mapping rule library, modifying the target flavor profile based on the flavor adjustment method to obtain a final flavor profile; collecting physical parameters of the coffee beans to be roasted in real time; inputting the physical parameters and the final flavor profile into a preset process parameter model to obtain a plurality of candidate process parameters, each candidate process parameter corresponding to a complete set of roasting control parameters; inputting each candidate process parameter and the physical parameters into a preset roasting result prediction model for processing to obtain a plurality of predicted flavor profiles, each predicted flavor profile corresponding to a candidate process parameter; calculating the similarity of each predicted flavor profile and the final flavor profile to obtain a plurality of quality similarity scores; selecting a maximum value from the plurality of quality similarity scores, determining the candidate process parameter corresponding to the maximum value as a baseline optimal process parameter, and sending the baseline optimal process parameter to the first user for viewing the optimal process parameter.

[0007] By adopting the above technical solution, the roasting requirement of the user is received and analyzed to accurately extract the flavor target and the brewing method. The flavor target is converted into quantifiable flavor indicators and scores based on the preset coffee flavor dimension table to construct the target flavor profile, realizing the standardized expression of the user's requirement. The target flavor profile is modified through the preset brewing flavor mapping rule library to obtain the final flavor profile, considering the influence of different brewing methods on the flavor. At the same time, the physical parameters of the coffee beans to be roasted are collected in real time, and the physical parameters and the final flavor profile are input into the preset process parameter model to generate a plurality of candidate process parameter schemes. The candidate process parameters and the physical parameters are input into the preset roasting result prediction model to predict the flavor profile corresponding to each scheme. The similarity of the predicted flavor profile and the final flavor profile is calculated to select the scheme with the highest similarity as the baseline optimal process parameter, breaking through the limitation of the traditional static preset roasting curve, realizing the dynamic generation of the optimal process parameter according to the specific raw material characteristics and the user's requirement, and more accurately matching the flavor characteristics expected by the user to improve the user experience.

[0008] Optionally, the target flavor image is corrected based on a flavor adjustment mode to obtain a final flavor image, specifically including: obtaining coffee attribute information of the to-be-roasted coffee beans, the coffee attribute information including any one of coffee bean variety information, coffee processing information, and coffee origin information; combining the brewing mode and the coffee attribute information to form a combination key, and inputting the combination key into a preset brewing flavor mapping rule library for querying to obtain a flavor correction vector, wherein the preset brewing flavor mapping rule library stores a mapping relationship between the combination key and the flavor correction vector; performing vector addition calculation on the target flavor image and the flavor correction vector to obtain a corrected flavor image; obtaining a score corresponding to a target flavor index from the corrected flavor image, and judging whether the score corresponding to the target flavor index is within a preset score range; when the score corresponding to the target flavor index is not within the preset score range, setting the score corresponding to the target flavor index as a target boundary value, the target boundary value being a value closest to the score corresponding to the target flavor index within the preset score range; when the score corresponding to the target flavor index is within the preset score range, retaining the score corresponding to the target flavor index; and combining all processed flavor indexes and corresponding scores to obtain the final flavor image.

[0009] By adopting the above technical solution, the variety, processing method, and origin of the to-be-roasted coffee beans are obtained, and the coffee attribute information is combined with the brewing mode to form a combination key, so that the corresponding flavor correction vector can be found in the preset brewing flavor mapping rule library. The combination key query mechanism ensures the accuracy and comprehensiveness of flavor correction. Vector addition operation is performed on the target flavor image and the flavor correction vector to obtain a corrected flavor image closer to the actual situation. The flavor index score exceeding the range is processed at the boundary, and the score closest to the target boundary value within the range is adjusted. The score within the range is retained. All processed flavor indexes and corresponding scores are combined to form a final flavor image, effectively improving the accuracy of roasting process parameter optimization.

[0010] Optionally, after selecting the maximum value from multiple quality similarity scores and determining the candidate process parameter corresponding to the maximum value as the baseline optimal process parameter, the method further includes: obtaining the actual bean temperature of the coffee beans to be roasted using the baseline optimal process parameter; determining the current roasting time point corresponding to the actual bean temperature based on the roasting start time point; retrieving the preset bean temperature-time curve from the baseline optimal process parameter and finding the target bean temperature corresponding to the current roasting time point from the preset bean temperature-time curve; calculating the difference between the actual bean temperature and the target bean temperature to obtain the real-time temperature deviation; determining the target adjustment amount based on the real-time temperature deviation, adjusting the air force or fire force in the baseline optimal process parameter based on the target adjustment amount to generate the real-time execution process parameter; and outputting the real-time execution process parameter to the roasting equipment controller so that the roasting equipment controller can roast based on the real-time execution process parameter.

[0011] By employing the above technical solution, the actual bean temperature during the roasting process is acquired in real time. Based on the start time of roasting, the current roasting time is determined. A preset bean temperature-time curve is retrieved from the baseline optimal process parameters. By finding the target bean temperature corresponding to the current roasting time, dynamic reference for the roasting process is achieved. The difference between the actual bean temperature and the target bean temperature is calculated to obtain the real-time temperature deviation, enabling timely detection of temperature anomalies during roasting. Based on the real-time temperature deviation, the target adjustment amount is determined, and the airflow or heat in the baseline optimal process parameters is adjusted accordingly, generating real-time execution process parameters that dynamically respond to temperature fluctuations during roasting. These real-time execution process parameters are output to the roasting equipment controller. Real-time monitoring overcomes the limitations of traditional fixed process parameters, effectively addressing various interference factors during roasting and ensuring the maximum fit between the actual roasting curve and the expected curve. This improves the accuracy of the roasting process and guarantees the flavor and quality of the final coffee product.

[0012] Optionally, the target adjustment amount is determined based on the real-time temperature deviation, specifically including: comparing the absolute value of the real-time temperature deviation with a preset deviation threshold, and determining the baseline adjustment amount based on the comparison result; calculating the difference between the real-time temperature deviation at the current baking time point and the historical temperature deviation at the previous baking time point to obtain the deviation change rate; comparing the deviation change rate with a preset positive change threshold and a preset negative change threshold to obtain a dynamic adjustment coefficient; multiplying the baseline adjustment amount with the dynamic adjustment coefficient to obtain the target adjustment amount, and determining the direction of action of the target adjustment amount based on the original sign of the real-time temperature deviation.

[0013] By adopting the above technical solution, the preset deviation threshold is compared with the absolute value of the real-time temperature deviation to determine the corresponding baseline adjustment amount. The deviation change rate is obtained by calculating the temperature deviation difference between the current baking time point and the previous time point, and compared with the preset positive and negative change thresholds to obtain the dynamic adjustment coefficient. The baseline adjustment amount is multiplied by the dynamic adjustment coefficient to obtain the target adjustment amount, and the adjustment direction is determined according to the sign of the real-time temperature deviation. This achieves precise control of firepower and airflow, taking into account not only the absolute magnitude of the temperature deviation but also the trend of deviation change. It can respond to temperature changes more quickly and accurately, avoid over-adjustment or under-adjustment, and improve the stability and control precision of the baking process.

[0014] Optionally, after outputting the real-time execution process parameters to the roasting equipment controller so that the roasting equipment controller can roast based on the real-time execution process parameters, the method further includes: receiving feedback information from a second user, the feedback information including user satisfaction score or finished product quality indicators; when the user satisfaction score is less than the preset satisfaction score, or the finished product quality indicators exceed the preset normal range, retrieving the actual process data record and baseline optimal process parameters corresponding to the target roasting, the target roasting being the coffee roasting corresponding to the feedback information; comparing and analyzing the actual process data record with the baseline optimal process parameters to obtain process parameter deviation segments; if the feedback information contains flavor defect information, searching for typical process deviation patterns matching the flavor defect information from the preset process deviation flavor library; matching the process parameter deviation segments with the typical process deviation patterns; if the process parameter deviation segments successfully match the typical process deviation patterns, identifying the process parameter deviation segments as an abnormal link, and generating a quality traceability report based on the abnormal link.

[0015] By adopting the above technical solution, feedback from consumers or quality inspectors is received, enabling objective evaluation of the roasting results. When user satisfaction scores fall below preset standards or quality indicators are abnormal, the system automatically retrieves relevant actual roasting process data records and baseline optimal process parameters for comparative analysis, thereby identifying process parameter deviation segments. By matching actual process parameter deviation segments with typical deviation patterns, the abnormal links leading to quality problems are accurately located. This allows for rapid identification of key process deviations affecting product quality and the generation of detailed quality traceability reports. This not only achieves precise traceability of roasting quality issues but also improves the controllability of the coffee roasting process and the stability of product quality.

[0016] Optionally, the physical parameters and the final flavor profile are input into a preset process parameter model to obtain multiple candidate process parameters. Specifically, this includes: integrating the physical parameters and the final flavor profile into an input feature vector; inputting the input feature vector into a reverse process recommendation model to obtain an initial process parameter set; using the final flavor profile as the optimization objective and the initial process parameter set as the starting point, using an optimization search algorithm to iteratively search within the preset process parameter space to generate an iterative candidate set; inputting the iterative candidate set and physical parameters into a baking result prediction model to obtain an initial predicted flavor profile; calculating the distance between the initial predicted flavor profile and the final flavor profile to obtain the objective function value; determining the next iteration direction of the optimization search algorithm based on the objective function value, until the optimization search algorithm is executed a preset number of times or the objective function value converges; and selecting the iterative candidate set with the smallest objective function value from all the initial process parameters evaluated during the iteration process as multiple candidate process parameters.

[0017] By adopting the above technical solution, physical parameters and the final flavor profile are integrated into an input feature vector, which is then input into a reverse process recommendation model. With the final flavor profile as the optimization objective, an optimization search algorithm is used to iteratively search within a preset process parameter space that includes all possible combinations of baking control parameters, generating an iterative candidate set. The iterative candidate set and physical parameters are then input into a baking result prediction model to obtain an initial predicted flavor profile and calculate the objective function value. The iterative direction of the optimization search is dynamically adjusted based on the objective function value until a preset number of iterations is reached or the objective function value converges. Finally, from all evaluated initial process parameter sets, a preset number of iterative candidate sets with the smallest objective function value are selected as candidate process parameters.

[0018] Optionally, the candidate process parameters and physical parameters are input into a preset roasting result prediction model for processing to obtain multiple predicted flavor profiles. Specifically, this includes: obtaining target candidate process parameters from multiple candidate process parameters, extracting a set of process feature parameters from the target candidate process parameters, and determining a process feature vector based on the process feature parameters; encoding the physical parameters of the coffee beans to be roasted to obtain a physical feature vector; concatenating the process feature vector with the physical feature vector to obtain a comprehensive input vector; inputting the comprehensive input vector into the preset roasting result prediction model to obtain a predicted flavor vector; combining the predicted flavor vector and the corresponding flavor index to form a predicted flavor profile corresponding to the target candidate process parameters; and processing all candidate process parameters to obtain multiple predicted flavor profiles.

[0019] By adopting the above technical solution, process feature parameters are extracted from candidate process parameters and process feature vectors are constructed. At the same time, physical parameters such as moisture content, density, size distribution, variety, and processing method of coffee beans are encoded into physical feature vectors. The process feature vector and the physical feature vector are concatenated to form a comprehensive input vector. The comprehensive input vector is input into a preset roasting result prediction model to accurately predict the corresponding flavor vector. The predicted flavor vector is combined with flavor indicators to form a predicted flavor profile. This process is repeated for all candidate process parameters to finally obtain a complete set of predicted flavor profiles, which significantly improves the accuracy of coffee roasting process parameter optimization.

[0020] A second aspect of this application provides a system for optimizing coffee production process parameters. The system includes a receiving unit, a first processing unit, a second processing unit, and a sending unit. The receiving unit receives the roasting requirements of a first user for coffee beans to be roasted, analyzes the roasting requirements, and obtains flavor targets and brewing methods. The first processing unit, based on a preset coffee flavor dimension table, quantifies the flavor targets into a set of flavor indicators and corresponding flavor scores to construct a target flavor profile. It searches for flavor adjustment methods corresponding to the brewing method from a preset brewing flavor mapping rule base, and corrects the target flavor profile based on the flavor adjustment methods to obtain the final flavor profile. It also collects physical parameters of the coffee beans to be roasted in real time. The second processing unit... The unit inputs physical parameters and the final flavor profile into a preset process parameter model to obtain multiple candidate process parameters, each corresponding to a complete set of baking control parameters. It then inputs each candidate process parameter and physical parameter into a preset baking result prediction model for processing, resulting in multiple predicted flavor profiles, each corresponding to a candidate process parameter. The unit calculates the similarity between each predicted flavor profile and the final flavor profile, obtaining multiple quality similarity scores. The sending unit selects the maximum value from the multiple quality similarity scores, determines the candidate process parameter corresponding to the maximum value as the baseline optimal process parameter, and sends the baseline optimal process parameter to the first user so that the first user can view the optimal process parameter.

[0021] In a third aspect, this application provides an electronic device including a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory, causing the electronic device to perform any of the methods described above in this application.

[0022] In a fourth aspect, this application provides a computer-readable storage medium storing instructions that, when executed, perform any of the methods described above in this application.

[0023] In summary, one or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. The system receives and analyzes user roasting requests, accurately extracting flavor targets and brewing methods. Based on a pre-set coffee flavor dimension table, it transforms flavor targets into quantifiable flavor indicators and scores, constructing a target flavor profile. This achieves a standardized expression of user needs. Through a pre-set brewing flavor mapping rule library, it considers the impact of different brewing methods on flavor and corrects the target flavor profile to obtain the final flavor profile. Simultaneously, it collects physical parameters of the coffee beans to be roasted in real time. Inputting these physical parameters and the final flavor profile into a pre-set process parameter model generates multiple candidate process parameter schemes. Inputting these candidate process parameters and physical parameters into a pre-set roasting result prediction model predicts the flavor profile corresponding to each scheme. By calculating the similarity between the predicted and final flavor profiles, the scheme with the highest similarity is selected as the baseline optimal process parameter. This overcomes the limitations of traditional static pre-set roasting curves, achieving the goal of dynamically generating optimal process parameters based on specific raw material characteristics and user needs. This allows for more accurate matching of the flavor characteristics expected by users, improving the user experience. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating a method for optimizing coffee production process parameters according to an embodiment of this application. Figure 2 This is a schematic diagram of the structure of a coffee production process parameter optimization system provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of this application.

[0025] Explanation of reference numerals in the attached drawings: 201, receiving unit; 202, first processing unit; 203, second processing unit; 204, transmitting unit; 300, electronic device; 301, processor; 302, memory; 303, user interface; 304, network interface; 305, communication bus. Detailed Implementation

[0026] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0027] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0028] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0029] Therefore, improving the user experience by overcoming the limitations of traditional static preset roasting curves is a pressing issue. This application provides a method for optimizing coffee production process parameters. Figure 1 This is a flowchart illustrating a method for optimizing coffee production process parameters according to an embodiment of this application. (Refer to...) Figure 1 The method includes the following steps S101-S108.

[0030] S101: Receives the roasting requirements of the first user for the coffee beans to be roasted, analyzes the roasting requirements, and obtains the flavor target and brewing method.

[0031] In step S101 above, a standardized input interface for roasting production personnel (the first user) is provided via a web interface or mobile application. This interface includes two core modules: a flavor target selection module and a brewing method setting module. In the flavor target selection module, the first user can select from a preset flavor option library, using standardized flavor descriptions such as "high body," "bright acidity," and "intense caramel sweetness," or precisely define the intensity of each flavor dimension using a slider. In the brewing method setting module, the first user needs to specify the expected brewing method for this batch of coffee beans, such as "pour-over," "espresso," or "cold brew." After receiving this input information, the selected flavor target and brewing method are then confirmed.

[0032] For example, when the first user selects the flavor objective of "highlighting jasmine aroma and citrus acidity" and specifies "hand-brewed" as the brewing method, the flavor objective and brewing method are determined through identification. This demand analysis method breaks through the limitations of traditional experience-based descriptions, achieving standardized expression and precise transmission of flavor requirements, laying the foundation for personalized customized production.

[0033] S102: Based on the preset coffee flavor dimension table, the flavor target is quantified into a set of flavor indicators and corresponding flavor scores to form a target flavor profile.

[0034] In S102 above, a complete preset coffee flavor dimension table was constructed in advance based on international coffee tasting standards. This dimension table covers key flavor dimensions such as aroma, acidity, sweetness, body, and finish. Each dimension is further subdivided into specific flavor descriptions; for example, the aroma dimension includes floral (jasmine, rose), fruity (citrus, berry), and caramel aromas. This dimension table was established and continuously optimized by combining a large amount of professional cupping data and empirical models.

[0035] The acquired flavor targets are mapped to a pre-defined coffee flavor dimension table, automatically converting qualitative flavor descriptions into standardized quantitative indicators. For example, when the flavor target is "highlighting jasmine aroma and citrus acidity," the floral aroma index under the aroma dimension is assigned a high score (e.g., 8.5 points), and the citrus acidity index under the acidity dimension is also assigned a high score (e.g., 8.0 points), while other dimensions such as body may be assigned a medium score (e.g., 6.0 points). These quantified flavor indicators and their corresponding scores together constitute a multi-dimensional target flavor profile, with each dimension's score falling within the standard range of 0-10. This quantitative method transforms abstract flavor requirements into measurable and comparable numerical indicators, accurately expressing and conveying flavor requirements. It effectively avoids misunderstandings that may arise from traditional empirical descriptions and provides a reliable evaluation benchmark for controlling the flavor consistency of different batches of products.

[0036] S103: Search for the flavor adjustment method corresponding to the brewing method from the preset brewing flavor mapping rule library, and modify the target flavor profile based on the flavor adjustment method to obtain the final flavor profile.

[0037] In step S103 above, a pre-established brewing flavor mapping rule library is used to address the impact of different brewing methods on coffee flavor expression. This pre-established rule library, based on extensive experimental data and professional cupping results, systematically analyzes and models the changes in coffee flavor under different brewing methods, recording the patterns of flavor characteristic changes under various brewing methods such as pour-over, espresso, and cold brew. Based on the brewing method determined by the first user, the corresponding flavor adjustment method is retrieved from the pre-established rule library. These adjustment methods are stored in the form of flavor correction vectors, containing the degree of enhancement or suppression of each flavor dimension under a specific brewing method. For example, for pour-over coffee, its extraction characteristics enhance the aroma and acidity of the coffee while somewhat weakening its body; therefore, the target flavor profile will be adjusted accordingly.

[0038] Furthermore, the target flavor profile is corrected based on flavor adjustment methods to obtain the final flavor profile. Specifically, this includes: acquiring the coffee attribute information of the coffee beans to be roasted, including any one of the following: coffee bean variety information, coffee processing information, and coffee origin information; constructing a key using the brewing method and coffee attribute information, and inputting this key into a preset brewing flavor mapping rule library for querying to obtain a flavor correction vector, where the preset brewing flavor mapping rule library stores the mapping relationship between the key and the flavor correction vector; performing vector addition on the target flavor profile and the flavor correction vector to obtain the corrected flavor profile; obtaining the score corresponding to the target flavor index from the corrected flavor profile, and determining whether the score corresponding to the target flavor index is within a preset score range; when the score corresponding to the target flavor index is not within the preset score range, setting the score corresponding to the target flavor index as the target boundary value, where the target boundary value is the value closest to the score corresponding to the target flavor index within the preset score range; when the score corresponding to the target flavor index is within the preset score range, retaining the score corresponding to the target flavor index; and combining all processed flavor indicators and their corresponding scores to obtain the final flavor profile.

[0039] Specifically, coffee attribute information of the coffee beans to be roasted is acquired through an IoT data acquisition layer or a manual input interface. This information includes the coffee bean variety (e.g., Ethiopian Yirgacheffe G1), processing method (e.g., washed, sun-dried), and origin information (e.g., altitude, region). This attribute information is closely related to the flavor potential of the coffee beans and has a significant impact on the final flavor expression. The acquired brewing method (e.g., pour-over) is concatenated with the coffee attribute information to form a unique key. For example, a key like "pour-over-Yirgacheffe G1-washed-high altitude" contains key factors affecting flavor expression. After this key is input into a preset brewing flavor mapping rule base, the corresponding flavor correction vector is retrieved. This rule base is built based on a large amount of experimental data and stores the flavor characteristics under different combination conditions. For example, for the above key, a flavor correction vector representing "aroma +1.2, acidity +0.8, sweetness -0.3, body -0.5" can be returned.

[0040] The obtained target flavor profile is then added to the retrieved flavor correction vector. If a flavor indicator (such as floral aroma) in the target flavor profile has a score of 8.5, and the corresponding correction vector value is +1.2, the corrected score will become 9.7. A range check is performed on each corrected flavor indicator score to determine if it falls within the preset score range of 0-10. When a corrected score is found to be outside the range (such as 9.7 in the example above), it is adjusted to the closest boundary value (10 in this case). This boundary handling ensures the practicality and feasibility of the flavor profile. Scores that remain within the 0-10 range after correction are retained, maintaining their accurate corrected values. All flavor indicators that have undergone range checking and boundary handling, along with their corresponding scores, are recombined to form the final flavor profile.

[0041] For example, the final flavor profile might be expressed as: "floral: 10.0, citrus acidity: 8.8, sweetness: 5.7, body: 5.5", etc. This comprehensively revised and standardized flavor profile takes into account both the characteristics of the coffee beans themselves and the influence of the brewing method, while ensuring that all indicators are within an achievable range, providing accurate and feasible target guidance for subsequent optimization of roasting process parameters.

[0042] S104: Real-time data collection of the coffee beans to be roasted to obtain physical parameters.

[0043] In step S104 above, multiple intelligent sensors and detection devices deployed on the production line collect comprehensive physical parameters of the coffee beans to be roasted. A high-precision moisture meter is used to detect the moisture content of the coffee beans in real time, a professional densitometer measures the density of the coffee beans, and a laser particle size analyzer obtains the size distribution data of the coffee beans. Simultaneously, barcode scanning or RFID technology is used to read the batch information of the coffee beans, retrieving the variety information (e.g., Arabica) and processing method information (e.g., washed, sun-dried, honey-processed) from the database. These detection devices communicate with the central control system in real time via industrial IoT protocols (e.g., OPC UA, MQTT) to ensure timely data transmission and synchronization.

[0044] For example, when a batch of Ethiopian Yirgacheffe G1 green coffee beans enters the processing stage, key physical parameters such as moisture content (e.g., 11.5%), density (e.g., 650 g / L), and size distribution (e.g., 90% of the beans are distributed in the 5.5-7.5 mm range) are collected and recorded in real time. These real-time collected physical parameters provide accurate raw material characteristic information for subsequent process parameter optimization, helping to precisely adjust roasting parameters according to the actual state of the raw materials, ensuring that the final product can fully demonstrate the excellent flavor characteristics of the coffee beans.

[0045] S105: Input the physical parameters and final flavor profile into the preset process parameter model to obtain multiple candidate process parameters.

[0046] In step S105 above, the physical parameters and the final flavor profile are input into a preset process parameter model to obtain multiple candidate process parameters. Specifically, this includes: integrating the physical parameters and the final flavor profile into an input feature vector; inputting the input feature vector into a reverse process recommendation model to obtain an initial process parameter set; using the final flavor profile as the optimization objective and the initial process parameter set as the starting point, using an optimization search algorithm to iteratively search within the preset process parameter space to generate an iterative candidate set; inputting the iterative candidate set and physical parameters into a baking result prediction model to obtain an initial predicted flavor profile; calculating the distance between the initial predicted flavor profile and the final flavor profile to obtain the objective function value; determining the next iteration direction of the optimization search algorithm based on the objective function value, until the optimization search algorithm is executed a preset number of times or the objective function value converges; and selecting the iterative candidate set with the smallest objective function value from all the initial process parameters evaluated during the iteration process as multiple candidate process parameters.

[0047] Specifically, the preset process parameter model is a machine learning model (such as Gradient Boosting Decision Tree (GBDT) or Deep Neural Network) pre-trained with a large amount of historical baking data. First, the acquired physical parameters (moisture content 11.5%, density 650 g / L, size distribution 90% between 5.5-7.5 mm, Yirgacheffe G1 variety, washed treatment) and the final flavor profile (floral aroma: 10.0, citrus acidity: 8.8, sweetness: 5.7, body: 5.5) are standardized and feature-encoded. Discrete parameters (such as variety and treatment method) are one-hot encoded, and continuous parameters (such as moisture content and density) are normalized. Finally, all features are integrated into a unified input feature vector.

[0048] The input feature vector is fed into a pre-trained reverse process recommendation model (using a deep neural network architecture, including multiple fully connected layers and dropout layers). This model is trained based on feature-parameter pairs from successful historical cases, enabling it to quickly generate an initial set of process parameters based on the input features. Each feature-parameter pair includes: the physical parameters of the coffee beans to be roasted, the target flavor profile (input), and the corresponding optimal roasting process parameter set (output). For example, the model might output a set of parameters: bean inlet temperature 180℃, first crack temperature 195℃, development time 2.5 minutes, etc. Starting with this initial set of process parameters, a Bayesian optimization algorithm is used to search within a pre-defined process parameter space. This parameter space contains all reasonable parameter combinations, such as bean inlet temperature range 160-200℃, first crack temperature range 185-205℃, development time range 1.5-4 minutes, etc. In each iteration, the optimization algorithm generates multiple candidate parameter combinations, forming an iterative candidate set. Each iterative candidate set, along with the original physical parameters, is input into a roasting result prediction model (trained using a GBDT model) to predict its potential flavor profile. This predictive model is trained based on the mapping relationship between 'actual roasting process parameters, coffee bean physical parameters' and 'flavor profile of the final roasted product' from a large amount of historical roasting data. For example, for a certain set of parameters, the predicted flavor profile might be: floral: 9.8, citrus acidity: 8.5, sweetness: 5.8, body: 5.6.

[0049] Then, the Euclidean distance between the initial predicted flavor profile and the target final flavor profile is calculated as the objective function value. The smaller the distance, the closer the prediction result is to the target flavor. For example, the distance between the above prediction result and the target flavor might be 0.35. Based on the calculated objective function value, the Gaussian process regression characteristic of the Bayesian optimization algorithm is used to predict the region in the parameter space most likely to reduce the objective function value, determining the search direction for the next iteration. This process continues until a preset number of 1000 iterations is reached, or the change in the objective function value is less than 0.01 for 50 consecutive iterations, at which point convergence is considered achieved. From all parameter combinations evaluated during the iteration process, the top 5 parameter combinations with the smallest objective function values ​​(a preset number) are selected as the final candidate process parameters. These parameter combinations can all achieve the target flavor well, but may employ different baking paths.

[0050] Each candidate process parameter includes a complete set of roasting control parameters, such as bean inlet temperature, roasting time, heating rate at different stages, first crack temperature, and final temperature. For example, for Yirgacheffe G1 green beans with a moisture content of 11.5%, the system may generate multiple sets of candidate parameters: the first set is "bean inlet temperature 180℃, first crack temperature 195℃, development time 2.5 minutes", the second set is "bean inlet temperature 175℃, first crack temperature 192℃, development time 3 minutes", etc.

[0051] S106: Input each candidate process parameter and physical parameter into the preset baking result prediction model for processing to obtain multiple predicted flavor profiles.

[0052] In step S106 above, each candidate process parameter and physical parameter is input into a preset roasting result prediction model for processing to obtain multiple predicted flavor profiles. Specifically, this includes: obtaining target candidate process parameters from multiple candidate process parameters, extracting a set of process feature parameters from the target candidate process parameters, and determining a process feature vector based on the process feature parameters; encoding the physical parameters of the coffee beans to be roasted to obtain a physical feature vector; concatenating the process feature vector with the physical feature vector to obtain a comprehensive input vector; inputting the comprehensive input vector into the preset roasting result prediction model to obtain a predicted flavor vector; combining the predicted flavor vector and the corresponding flavor index to form a predicted flavor profile corresponding to the target candidate process parameters; and processing all candidate process parameters to obtain multiple predicted flavor profiles.

[0053] Specifically, each set of process parameters is selected sequentially from multiple candidate process parameters as target candidate process parameters for processing. For each set of target candidate process parameters, key process feature parameters are extracted, including bean inlet temperature, roasting time, heating rate at different stages, first crack temperature, and final temperature. These parameters are standardized and converted into numerical values ​​with uniform dimensions, and then organized into process feature vectors according to a preset feature template. For example, when processing the set of parameters "bean inlet temperature 180℃, first crack temperature 195℃, development time 2.5 minutes", it is converted into a standardized feature vector form [0.75, 0.82, 0.63...].

[0054] Simultaneously, physical parameters are encoded. For continuous parameters such as moisture content (11.5%), density (650 g / L), and size distribution (90% within 5.5-7.5 mm), a min-max normalization method is used for standardization. For discrete parameters such as varietal information (Yirgacheffe G1) and processing method information (washed method), one-hot encoding is used to convert them into numerical vectors. All these encoded features are combined into a physical feature vector [0.58, 0.65, 0.72, 1, 0, 1, 0...]. The processing feature vector and the physical feature vector are concatenated in a preset order to construct a complete integrated input vector. This vector contains information on all the key factors affecting coffee flavor formation, providing complete input data for subsequent flavor prediction.

[0055] The constructed comprehensive input vector is fed into the preset baking result prediction model. This model is a regression model based on a deep neural network architecture, containing multiple fully connected layers and dropout layers, and has been trained using over 5000 batches of historical baking data. The preset baking result prediction model accurately captures the complex nonlinear relationship between process parameters, physical properties, and the final flavor. The model outputs a predicted flavor vector, where each dimension corresponds to a specific flavor index prediction score.

[0056] For example, given the input above, the model might output a vector [9.8, 8.5, 5.8, 5.6], corresponding to predicted scores for floral aroma, citrus acidity, sweetness, and body, respectively. These predicted scores are then paired with their corresponding flavor index names to construct a complete predicted flavor profile: "Floral aroma: 9.8, Citrus acidity: 8.5, Sweetness: 5.8, Body: 5.6". This predicted flavor profile visually demonstrates the flavor characteristics that coffee might exhibit under specific process parameters and physical conditions. Repeating the above process for all candidate process parameters yields multiple predicted flavor profiles.

[0057] S107: Calculate the similarity between each predicted flavor profile and the final flavor profile to obtain multiple quality similarity scores.

[0058] In S107 above, a similarity analysis is performed between each predicted flavor profile and the final flavor profile. First, the flavor profile is converted into a standardized flavor vector form, where each dimension corresponds to a score of a flavor index. For example, a predicted flavor profile "floral: 9.8, citric acid: 8.5, sweetness: 5.8, body: 5.6" is converted into a vector [9.8, 8.5, 5.8, 5.6], and the final flavor profile "floral: 10.0, citric acid: 8.8, sweetness: 5.7, body: 5.5" is converted into a vector [10.0, 8.8, 5.7, 5.5].

[0059] A weighted cosine similarity algorithm is used to calculate similarity scores, assigning differentiated weights to different flavor dimensions. Key flavor indicators such as floral notes and citrus acidity have weights of 0.4 and 0.3, respectively, while secondary flavor indicators such as sweetness and body have weights of 0.2 and 0.1, respectively. A standardized quality similarity score is obtained by calculating the cosine of the angle between the weighted vectors. The weighted cosine similarity algorithm effectively measures the directional similarity between multi-dimensional vectors and, by assigning higher weights to key flavor dimensions, makes them more consistent with human sensory preferences for coffee flavors. For example, the quality similarity score between predicted flavor profile A and the final flavor profile is 0.997, while the quality similarity score between predicted flavor profile B and the final flavor profile is 0.995. This weighted cosine similarity calculation yields a quality similarity score between 0 and 1; the closer the score is to 1, the more similar the predicted flavor is to the target flavor.

[0060] S108: Select the maximum value from multiple quality similarity scores, determine the candidate process parameter corresponding to the maximum value as the baseline optimal process parameter, and send the baseline optimal process parameter to the first user so that the first user can view the optimal process parameter.

[0061] In step S108 above, after calculating multiple quality similarity scores, all quality similarity scores are filtered. A quicksort algorithm can be used to sort the multiple quality similarity scores in descending order, and the maximum value is selected. For example, when comparing multiple sets of data such as "process parameter A - similarity 0.997", "process parameter B - similarity 0.995", and "process parameter C - similarity 0.991", the maximum value of 0.997 is selected. Then, the candidate process parameters corresponding to this maximum similarity score are obtained and determined as the baseline optimal process parameters. This set of baseline optimal process parameters includes a complete set of roasting control parameters, such as "bean inlet temperature 180℃, first crack temperature 195℃, development time 2.5 minutes", etc.

[0062] This set of baseline optimal process parameters is pushed to the production management system's interface in real time via industrial IoT communication protocols (such as OPC UA). Simultaneously, push notifications are sent to the terminal devices of the primary user (such as the baking process manager) through the enterprise's internal instant messaging system. The primary user can view detailed process parameter information, including parameter values ​​and expected flavor effects, via mobile terminal or workstation. This automated parameter optimization and push mechanism not only improves the efficiency of process optimization but also ensures that key process information is delivered to relevant personnel in a timely and accurate manner, providing strong support for achieving precise quality control.

[0063] In one possible implementation, after generating baseline optimal process parameters, the roasting equipment is controlled to roast the coffee beans according to these parameters. During roasting, the actual temperature of the coffee beans is monitored in real time. When the actual temperature deviates from the baseline optimal process parameters, the system effectively compensates for roasting deviations caused by differences in ambient temperature, raw material characteristics, etc., ensuring that the actual roasting process strictly follows the preset optimal process route. Real-time monitoring during roasting includes: obtaining the actual bean temperature of the coffee beans roasted using the baseline optimal process parameters; determining the current roasting time point corresponding to the actual bean temperature based on the roasting start time point; retrieving the preset bean temperature-time curve from the baseline optimal process parameters and finding the target bean temperature corresponding to the current roasting time point from the preset bean temperature-time curve; calculating the difference between the actual bean temperature and the target bean temperature to obtain the real-time temperature deviation; determining the target adjustment amount based on the real-time temperature deviation; adjusting the airflow or heat in the baseline optimal process parameters based on the target adjustment amount to generate real-time execution process parameters; and outputting the real-time execution process parameters to the roasting equipment controller so that the roasting equipment controller can roast based on the real-time execution process parameters.

[0064] Specifically, high-precision temperature sensors installed in the roasting equipment collect real-time data on the actual temperature of the coffee beans to be roasted. These sensors use PT100 platinum resistance temperature sensors, sampling at a frequency of 10 times per second to ensure the accuracy and real-time nature of the temperature data. For example, the actual bean temperature collected at a certain moment is 175.8℃. By recording the timestamp of the roasting start, the time difference between the current sampling time and the roasting start time is calculated to determine the current roasting time. For example, if the roasting start time is 14:30:00 and the current time is 14:35:30, then the current roasting time is determined to be 5 minutes and 30 seconds.

[0065] The preset bean temperature-time curve data is retrieved from the baseline optimal process parameters. This curve is a continuous curve formed by piecewise cubic spline interpolation of temperature values ​​at multiple key time points, recording the target temperature to be reached at each time point throughout the entire roasting process. Based on the current roasting time (5 minutes and 30 seconds), the corresponding target bean temperature is found in the preset curve; for example, the target bean temperature is found to be 176.5℃. The difference between the actual collected bean temperature of 175.8℃ and the target bean temperature of 176.5℃ is calculated, yielding a real-time temperature deviation of -0.7℃. This deviation indicates that the current actual roasting temperature is slightly lower than the expected target.

[0066] In addition, the target adjustment amount is determined based on the real-time temperature deviation, specifically including: comparing the absolute value of the real-time temperature deviation with a preset deviation threshold, and determining the baseline adjustment amount based on the comparison result; calculating the difference between the real-time temperature deviation at the current baking time point and the historical temperature deviation at the previous baking time point to obtain the deviation change rate; comparing the deviation change rate with a preset positive change threshold and a preset negative change threshold to obtain a dynamic adjustment coefficient; multiplying the baseline adjustment amount with the dynamic adjustment coefficient to obtain the target adjustment amount, and determining the direction of action of the target adjustment amount based on the original sign of the real-time temperature deviation.

[0067] Specifically, the absolute value of the current real-time temperature deviation (-0.7℃) is first obtained (0.7℃), and then compared with a preset deviation threshold. In this embodiment, the preset deviation threshold includes a first preset deviation threshold and a second preset deviation threshold. The absolute value of the current real-time temperature deviation is compared with both the first and second preset deviation thresholds, and a baseline adjustment amount is determined based on the comparison result. If the absolute value of the current real-time temperature deviation is greater than the first preset deviation threshold but less than the second preset deviation threshold, it is determined to be a medium deviation level, and the baseline adjustment amount corresponding to the medium deviation level can be set to 3%. If the absolute value of the current real-time temperature deviation is less than the first preset deviation threshold, the baseline adjustment amount is set to 1%. If the absolute value of the current real-time temperature deviation is greater than the second preset deviation threshold, the baseline adjustment amount is set to 5%. For example, the first preset deviation threshold can be set to 0.5℃, the second preset deviation threshold can be set to 1℃, and the absolute value of the current real-time temperature deviation is 0.7℃. In this case, it is greater than the first preset deviation threshold but less than the second preset deviation threshold, so the baseline adjustment amount is 3%.

[0068] Next, obtain the historical temperature deviation from the previous sampling time point, and calculate the difference between the previous sampling time point and the current real-time temperature deviation to obtain the deviation change rate. Compare this deviation change rate with preset positive change thresholds and preset negative change thresholds. If the deviation change rate is greater than the preset positive change threshold, set the dynamic adjustment coefficient to a first value greater than 1; if the deviation change rate is less than the preset negative change threshold, set the dynamic adjustment coefficient to a second value less than 1; if the deviation change rate is between the preset negative change threshold and the preset positive change threshold, set the dynamic adjustment coefficient to 1.

[0069] For example, if the previous sampling time point was 5 seconds ago and the historical temperature deviation was -0.5℃, the difference between the historical temperature deviation and the current real-time temperature deviation of -0.7℃ is calculated, yielding a deviation change rate of -0.04℃ / second. This deviation change rate is compared with preset positive change thresholds of 0.05℃ / second and preset negative change thresholds of -0.05℃ / second. Since -0.04℃ / second falls between these two thresholds, it indicates a relatively gentle trend in temperature deviation, and the dynamic adjustment coefficient is set to 1. If the deviation change rate exceeds 0.05℃ / second, it indicates a rapid increase in temperature deviation, and the dynamic adjustment coefficient is set to 1.5 to accelerate the response; if it is below -0.05℃ / second, it indicates a rapid convergence of temperature deviation, and the dynamic adjustment coefficient is set to 0.8 to avoid over-adjustment.

[0070] The target adjustment is obtained by multiplying the baseline adjustment by the dynamic adjustment coefficient. For example, if the baseline adjustment is 3%, the dynamic adjustment coefficient is 1, and the target adjustment is 3%, then since the original temperature deviation is negative (-0.7℃), indicating that the actual temperature is lower than the target temperature, this 3% adjustment is determined to be applied to increase firepower or decrease wind speed. If the current firepower is 60%, the adjusted firepower will increase to 63%; or if the current wind speed is 80%, the adjusted wind speed will decrease to 77%. By combining an adaptive control strategy of static grading and dynamic rate of change, the system's timely response to temperature deviations is ensured while avoiding potential oscillations during control.

[0071] Furthermore, after determining the target adjustment amount, the calculated adjustment parameters are integrated with the original baseline optimal process parameters to generate new real-time execution process parameters. For example, if the firepower in the original baseline parameters is 60%, based on the 3% adjustment amount, the firepower value in the new real-time execution process parameters will be adjusted to 63%.

[0072] Real-time process parameters are sent to the programmable logic controller (PLC) of the baking equipment via an industrial fieldbus (such as Modbus TCP). Upon receiving the new parameters, the PLC immediately adjusts the output of the actuators (such as variable frequency fans and proportional control valves) to achieve precise control of the baking process.

[0073] In one possible implementation, data-driven product traceability can quickly pinpoint key process deviations affecting product quality, providing precise improvement directions for production optimization. By establishing a correlation analysis between process deviations and flavor defects, not only can problems be quickly located, but similar problems can also be prevented from recurring. Specifically, this includes: receiving feedback information from a second user, which includes user satisfaction scores or finished product quality indicators; when the user satisfaction score is lower than a preset satisfaction score, or the finished product quality indicators exceed a preset normal range, retrieving the actual process data record and baseline optimal process parameters corresponding to the target roast (the coffee roast corresponding to the feedback information); comparing and analyzing the actual process data record with the baseline optimal process parameters to obtain process parameter deviation segments; if the feedback information contains flavor defect information, searching for typical process deviation patterns matching the flavor defect information from a preset process deviation flavor library; matching the process parameter deviation segments with the typical process deviation patterns; if the process parameter deviation segments successfully match the typical process deviation patterns, the process parameter deviation segments are identified as an abnormal link, and a quality traceability report is generated based on the abnormal link.

[0074] Specifically, feedback information submitted by a second user (e.g., a consumer or quality control personnel) is received through the user feedback interface of the enterprise-level digital management platform. This feedback information includes standardized user satisfaction scores or finished product quality indicator data measured by professional testing instruments. The received user satisfaction scores are compared with preset satisfaction benchmarks, or the finished product quality indicators are compared with preset normal ranges for quality parameters. When a score is found to be below the benchmark or an indicator exceeds the range, the quality traceability process is automatically triggered.

[0075] For example, when quality inspectors find that a batch of coffee products has a user satisfaction rating of 7.5 (lower than the preset satisfaction rating of 8.5), or detect that the acidity value of the coffee product is 8.9 (outside the preset normal range of 5.0-8.5), and the feedback includes flavor defect information such as "obvious burnt taste", the system will immediately initiate the quality traceability process.

[0076] The complete actual process data record for the corresponding batch of coffee is retrieved from the production management database, including time-series data of process parameters such as temperature, heat, and airflow throughout the roasting process. Simultaneously, the corresponding baseline optimal process parameters are retrieved as a comparison standard. Data analysis algorithms are used to compare the actual process data record and the baseline optimal process parameters over time. By calculating the parameter deviation values ​​at each time point and combining them with preset deviation judgment rules, process parameter segments that significantly deviate from the baseline are identified and extracted as process parameter deviation segments. When the feedback information contains a description of flavor defects, a preset process deviation flavor library can be accessed. This flavor library stores a large number of correspondence models between different process deviations in historical production processes and flavor defects in the final product. Natural language processing technology is used to parse the flavor defect information and retrieve and match corresponding typical process deviation patterns.

[0077] Pattern recognition algorithms are used to match extracted process parameter deviation segments with retrieved typical process deviation patterns. The degree of matching between the process parameter deviation segments and typical patterns is evaluated by calculating time series similarity. When the similarity exceeds a preset matching threshold, a successful match is determined, and the process parameter deviation segment is marked as an abnormal step affecting product quality. A quality traceability report is automatically generated based on the confirmed abnormal step. The report records detailed information such as the time location of the abnormal step, parameter deviation characteristics, and degree of impact, and provides possible cause analysis and improvement suggestions based on a process expert knowledge base. This data-driven quality traceability mechanism can quickly and accurately locate key process deviations affecting product quality, providing precise improvement directions for production process optimization.

[0078] For example, by comparing the actual process data records with the baseline optimal process parameters, it was found that at 8 minutes of roasting, the actual bean temperature exceeded the baseline curve by 2.5℃, lasting for about 45 seconds. This deviation record was extracted as a process parameter deviation segment. When the feedback information contained flavor defect information of "obvious burnt taste," a query was performed on the preset process deviation flavor library, and typical process deviation patterns related to "burnt taste" were retrieved: a positive temperature deviation of more than 2℃ before and after the first crack, lasting for more than 30 seconds. By calculating the similarity score of the two time series, which was 0.92, which was greater than the preset matching threshold of 0.85, it was confirmed that the process parameter deviation segment successfully matched the typical process deviation pattern. This indicates that a typical process deviation that could lead to a burnt taste did indeed occur during the roasting process of this batch. The report recorded in detail the specific time period of the anomaly (8 minutes 00 seconds to 8 minutes 45 seconds), the degree of deviation (maximum temperature deviation of 2.5℃), the duration (45 seconds), and other key information, as well as possible cause analysis (such as excessively rapid temperature rise before the first crack) and improvement suggestions (such as improving temperature control accuracy before the first crack).

[0079] This application also provides a system for optimizing coffee production process parameters. Figure 2 This is a schematic diagram of a system for optimizing coffee production process parameters provided in an embodiment of this application. (Refer to...) Figure 2 The system includes a receiving unit 201, a first processing unit 202, a second processing unit 203, and a transmitting unit 204. The receiving unit 201 receives the roasting requirements of the first user for the coffee beans to be roasted, analyzes the roasting requirements, and obtains the flavor target and brewing method. The first processing unit 202 quantifies the flavor target into a set of flavor indicators and corresponding flavor scores based on a preset coffee flavor dimension table to form a target flavor profile; it searches for flavor adjustment methods corresponding to the brewing method from a preset brewing flavor mapping rule library, and corrects the target flavor profile based on the flavor adjustment methods to obtain the final flavor profile; it collects physical parameters of the coffee beans to be roasted in real time. The second processing unit 203 inputs the physical parameters and the final flavor profile into a preset process parameter model to obtain multiple candidate process parameters, each candidate process parameter corresponding to a complete set of baking control parameters; inputs each candidate process parameter and physical parameter into a preset baking result prediction model for processing to obtain multiple predicted flavor profiles, each predicted flavor profile corresponding to a candidate process parameter; and calculates the similarity between each predicted flavor profile and the final flavor profile to obtain multiple quality similarity scores. The sending unit 204 selects the maximum value from multiple quality similarity scores, determines the candidate process parameter corresponding to the maximum value as the baseline optimal process parameter, and sends the baseline optimal process parameter to the first user so that the first user can view the optimal process parameter.

[0080] In one possible implementation, the receiving unit 201 is used to acquire coffee attribute information of the coffee beans to be roasted, including any one of coffee bean variety information, coffee processing information, and coffee origin information; the first processing unit 202 is used to construct a combination key by combining the brewing method and coffee attribute information, and input the combination key into a preset brewing flavor mapping rule library for querying to obtain a flavor correction vector, wherein the preset brewing flavor mapping rule library stores the mapping relationship between the combination key and the flavor correction vector; the target flavor profile and the flavor correction vector are calculated by vector addition to obtain the corrected flavor profile; the score corresponding to the target flavor index is obtained from the corrected flavor profile, and it is determined whether the score corresponding to the target flavor index is within a preset score range; when the score corresponding to the target flavor index is not within the preset score range, the score corresponding to the target flavor index is set as the target boundary value, and the target boundary value is the value closest to the score corresponding to the target flavor index within the preset score range; when the score corresponding to the target flavor index is within the preset score range, the score corresponding to the target flavor index is retained; all processed flavor indicators and their corresponding scores are combined to obtain the final flavor profile.

[0081] In one possible implementation, the receiving unit 201 is used to acquire the actual bean temperature of the coffee beans to be roasted using baseline optimal process parameters; the second processing unit 203 is used to determine the current roasting time point corresponding to the actual bean temperature based on the roasting start time point; retrieve a preset bean temperature-time curve from the baseline optimal process parameters, and find the target bean temperature corresponding to the current roasting time point from the preset bean temperature-time curve; calculate the difference between the actual bean temperature and the target bean temperature to obtain the real-time temperature deviation; determine the target adjustment amount based on the real-time temperature deviation, and adjust the air force or fire force in the baseline optimal process parameters based on the target adjustment amount to generate real-time execution process parameters; the sending unit 204 is used to output the real-time execution process parameters to the roasting equipment controller so that the roasting equipment controller can roast based on the real-time execution process parameters.

[0082] In one possible implementation, the second processing unit 203 is used to compare the absolute value of the real-time temperature deviation with a preset deviation threshold, and determine the baseline adjustment amount based on the comparison result; calculate the difference between the real-time temperature deviation at the current baking time point and the historical temperature deviation at the previous baking time point to obtain the deviation change rate; compare the deviation change rate with a preset positive change threshold and a preset negative change threshold to obtain a dynamic adjustment coefficient; multiply the baseline adjustment amount with the dynamic adjustment coefficient to obtain the target adjustment amount, and determine the direction of action of the target adjustment amount based on the original sign of the real-time temperature deviation.

[0083] In one possible implementation, the receiving unit 201 is used to receive feedback information from a second user, which includes a user satisfaction score or a finished product quality indicator; the second processing unit 203 is used to retrieve the actual process data record and baseline optimal process parameters corresponding to the target roasting when the user satisfaction score is less than a preset satisfaction score or the finished product quality indicator exceeds a preset normal range, wherein the target roasting is the coffee roasting corresponding to the feedback information; compare and analyze the actual process data record and the baseline optimal process parameters to obtain a process parameter deviation segment; if the feedback information contains flavor defect information, search for a typical process deviation pattern that matches the flavor defect information from a preset process deviation flavor library; match the process parameter deviation segment with the typical process deviation pattern; if the process parameter deviation segment matches the typical process deviation pattern successfully, determine the process parameter deviation segment as an abnormal link, and generate a quality traceability report based on the abnormal link.

[0084] In one possible implementation, the second processing unit 203 is used to integrate physical parameters and the final flavor profile into an input feature vector; input the input feature vector into a reverse process recommendation model to obtain an initial process parameter set; take the final flavor profile as the optimization objective and the initial process parameter set as the starting point, and use an optimization search algorithm to perform an iterative search within a preset process parameter space to generate an iterative candidate set; input the iterative candidate set and physical parameters into a baking result prediction model to obtain an initial predicted flavor profile; calculate the distance between the initial predicted flavor profile and the final flavor profile to obtain the objective function value; determine the next iteration direction of the optimization search algorithm based on the objective function value, until the optimization search algorithm is executed a preset number of times or the objective function value converges; and select the iterative candidate set with the smallest objective function value from all the initial process parameters evaluated during the iteration process as multiple candidate process parameters.

[0085] In one possible implementation, the receiving unit 201 is used to obtain target candidate process parameters from multiple candidate process parameters, extract a set of process feature parameters from the target candidate process parameters, and determine a process feature vector based on the process feature parameters; the second processing unit 203 is used to encode the physical parameters of the coffee beans to be roasted to obtain a physical feature vector; concatenate the process feature vector with the physical feature vector to obtain a comprehensive input vector; input the comprehensive input vector into a preset roasting result prediction model to obtain a predicted flavor vector; combine the predicted flavor vector and the corresponding flavor index to form a predicted flavor profile corresponding to the target candidate process parameters; and process all candidate process parameters to obtain multiple predicted flavor profiles.

[0086] It should be noted that the system provided in the above embodiments is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0087] This application also discloses an electronic device. (See reference...) Figure 3 , Figure 3 This application provides a schematic diagram of the structure of an electronic device. The electronic device 300 may include: at least one processor 301, at least one network interface 304, a user interface 303, a memory 302, and at least one communication bus 305.

[0088] The communication bus 305 is used to enable communication between these components.

[0089] The user interface 303 may include a display screen and a camera. Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0090] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0091] The processor 301 may include one or more processing cores. The processor 301 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 302, and by calling data stored in memory 302. Optionally, the processor 301 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 301 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and application requests; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 301 and may be implemented as a separate chip.

[0092] The memory 302 may include random access memory (RAM) or read-only memory. Optionally, the memory 302 may include a non-transitory computer-readable storage medium. The memory 302 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 302 may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch functionality, sound playback functionality, image playback functionality, etc.), instructions for implementing the various method embodiments described above, etc. The data storage area may store data involved in the various method embodiments described above. Optionally, the memory 302 may also be at least one storage device located remotely from the aforementioned processor 301.

[0093] like Figure 3 As shown, the memory 302, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for optimizing coffee production process parameters.

[0094] exist Figure 3In the electronic device 300 shown, the user interface 303 is mainly used to provide an interface for users to input data and obtain user input data; while the processor 301 can be used to call the application program stored in the memory 302 that optimizes coffee production process parameters. When executed by one or more processors, the electronic device performs one or more of the methods described in the above embodiments.

[0095] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0096] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0097] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some service interfaces; indirect couplings or communication connections between devices or units may be electrical or other forms.

[0098] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0099] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0100] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0101] The above description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of other embodiments of this disclosure upon considering the specification and the disclosure of practical truths. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure.

Claims

1. A method for optimizing coffee production process parameters, characterized in that, The method includes: Receive the roasting requirements of the first user for the coffee beans, analyze the roasting requirements, and obtain the flavor target and brewing method; Based on a preset coffee flavor dimension table, the flavor target is quantified into a set of flavor indicators and corresponding flavor scores to form a target flavor profile. The flavor adjustment method corresponding to the brewing method is searched in the preset brewing flavor mapping rule library, and the target flavor profile is modified based on the flavor adjustment method to obtain the final flavor profile; The physical parameters of the coffee beans to be roasted are collected in real time. The physical parameters and the final flavor profile are input into a preset process parameter model to obtain multiple candidate process parameters. Each candidate process parameter corresponds to a complete set of baking control parameters. Each of the candidate process parameters and the physical parameters is input into a preset baking result prediction model for processing to obtain multiple predicted flavor profiles, with each predicted flavor profile corresponding to one of the candidate process parameters. The similarity between each predicted flavor profile and the final flavor profile is calculated to obtain multiple quality similarity scores; The maximum value is selected from multiple quality similarity scores, and the candidate process parameter corresponding to the maximum value is determined as the baseline optimal process parameter. The baseline optimal process parameter is then sent to the first user so that the first user can view the optimal process parameter.

2. The method according to claim 1, characterized in that, The step of refining the target flavor profile based on the flavor adjustment method to obtain the final flavor profile specifically includes: Obtain the coffee attribute information of the coffee beans to be roasted, wherein the coffee attribute information includes any one of the following: coffee bean variety information, coffee processing information, and coffee origin information; The brewing method and the coffee attribute information are combined into a key, and the key is entered into the preset brewing flavor mapping rule library for querying to obtain a flavor correction vector. The preset brewing flavor mapping rule library stores the mapping relationship between the key and the flavor correction vector. The target flavor profile and the flavor correction vector are added together to obtain the corrected flavor profile. Obtain the score corresponding to the target flavor index from the corrected flavor profile, and determine whether the score corresponding to the target flavor index is within the preset score range; When the score corresponding to the target flavor index is not within the preset score range, the score corresponding to the target flavor index is set as the target boundary value, and the target boundary value is the value within the preset score range that is closest to the score corresponding to the target flavor index. When the score corresponding to the target flavor index is within the preset score range, the score corresponding to the target flavor index is retained. The final flavor profile is obtained by combining all the processed flavor indicators and their corresponding scores.

3. The method according to claim 1, characterized in that, After selecting the maximum value from multiple quality similarity scores and determining the candidate process parameter corresponding to the maximum value as the baseline optimal process parameter, the method further includes: The actual bean temperature at which the coffee beans to be roasted are roasted using the baseline optimal process parameters is obtained. The current baking time point corresponding to the actual bean temperature is determined based on the baking start time point; Retrieve the preset bean temperature-time curve from the baseline optimal process parameters, and find the target bean temperature corresponding to the current roasting time point from the preset bean temperature-time curve; The difference between the actual bean temperature and the target bean temperature is calculated to obtain the real-time temperature deviation; The target adjustment amount is determined based on the real-time temperature deviation, and the wind force or fire force in the baseline optimal process parameters is adjusted based on the target adjustment amount to generate real-time execution process parameters. The real-time execution process parameters are output to the baking equipment controller so that the baking equipment controller can perform baking based on the real-time execution process parameters.

4. The method according to claim 3, characterized in that, The step of determining the target adjustment amount based on the real-time temperature deviation specifically includes: The absolute value of the real-time temperature deviation is compared with a preset deviation threshold, and the benchmark adjustment amount is determined based on the comparison result. The difference between the real-time temperature deviation at the current baking time point and the historical temperature deviation at the previous baking time point is calculated to obtain the deviation change rate. The deviation change rate is compared with a preset positive change threshold and a preset negative change threshold to obtain a dynamic adjustment coefficient; The target adjustment amount is obtained by multiplying the baseline adjustment amount by the dynamic adjustment coefficient, and the direction of action of the target adjustment amount is determined according to the original sign of the real-time temperature deviation.

5. The method according to claim 3, characterized in that, After outputting the real-time execution process parameters to the baking equipment controller so that the baking equipment controller performs baking based on the real-time execution process parameters, the method further includes: Receive feedback information from a second user, the feedback information including user satisfaction rating or finished product quality indicators; When the user satisfaction score is less than the preset satisfaction score, or the finished product quality index exceeds the preset normal range, the actual process data record corresponding to the target roasting and the baseline optimal process parameters are retrieved, and the target roasting is the coffee roasting corresponding to the feedback information; The actual process data records are compared and analyzed with the baseline optimal process parameters to obtain process parameter deviation segments; If the feedback information contains flavor defect information, then a typical process deviation pattern that matches the flavor defect information is searched from the preset process deviation flavor library. Match the process parameter deviation segment with the typical process deviation pattern; If the process parameter deviation segment successfully matches the typical process deviation pattern, the process parameter deviation segment is identified as an abnormal link, and a quality traceability report is generated based on the abnormal link.

6. The method according to claim 1, characterized in that, The process involves inputting the physical parameters and the final flavor profile into a preset process parameter model to obtain multiple candidate process parameters, specifically including: The physical parameters and the final flavor profile are integrated into an input feature vector; The input feature vector is input into the reverse process recommendation model to obtain the initial process parameter set; Taking the final flavor profile as the optimization target and the initial process parameter set as the starting point, an optimization search algorithm is used to perform iterative search within the preset process parameter space to generate an iterative candidate set. The iterative candidate set and the physical parameters are input into the baking result prediction model to obtain an initial predicted flavor profile; Calculate the distance between the initial predicted flavor profile and the final flavor profile to obtain the objective function value; The next iteration direction of the optimization search algorithm is determined based on the objective function value. After the optimization search algorithm is executed a preset number of times or the objective function value converges, a preset number of iteration candidate sets with the smallest objective function value are selected from all the initial process parameters evaluated during the iteration process as the plurality of candidate process parameters.

7. The method according to claim 1, characterized in that, The step of inputting each of the candidate process parameters and the physical parameters into a preset baking result prediction model for processing to obtain multiple predicted flavor profiles specifically includes: A target candidate process parameter is obtained from a plurality of candidate process parameters, and a set of process feature parameters is extracted from the target candidate process parameter. A process feature vector is determined based on the process feature parameters. The physical parameters of the coffee beans to be roasted are encoded to obtain a physical feature vector; The process feature vector and the physical feature vector are concatenated to obtain the comprehensive input vector; The comprehensive input vector is input into the preset baking result prediction model to obtain the predicted flavor vector; The predicted flavor vector and the corresponding flavor index are combined to form the predicted flavor profile corresponding to the target candidate process parameters; All the candidate process parameters are processed to obtain multiple predicted flavor profiles.

8. A system for optimizing coffee production process parameters, characterized in that, The system includes a receiving unit, a first processing unit, a second processing unit, and a sending unit. The receiving unit receives the roasting requirements of the first user for the coffee beans to be roasted, analyzes the roasting requirements, and obtains the flavor target and brewing method. The first processing unit quantifies the flavor target into a set of flavor indicators and corresponding flavor scores based on a preset coffee flavor dimension table to form a target flavor profile. The flavor adjustment method corresponding to the brewing method is found in the preset brewing flavor mapping rule library, and the target flavor profile is modified based on the flavor adjustment method to obtain the final flavor profile. The physical parameters of the coffee beans to be roasted are collected in real time. The second processing unit inputs the physical parameters and the final flavor profile into a preset process parameter model to obtain multiple candidate process parameters, and each candidate process parameter corresponds to a complete set of baking control parameters; Each of the candidate process parameters and the physical parameters is input into a preset baking result prediction model for processing to obtain multiple predicted flavor profiles, with each predicted flavor profile corresponding to one of the candidate process parameters; the similarity between each predicted flavor profile and the final flavor profile is calculated to obtain multiple quality similarity scores; The sending unit selects the maximum value from multiple quality similarity scores, determines the candidate process parameter corresponding to the maximum value as the baseline optimal process parameter, and sends the baseline optimal process parameter to the first user so that the first user can view the optimal process parameter.

9. An electronic device, characterized in that, The device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1-7.

Images